Haven't measured the time, but I have seen a pause before commit happens. As stated during my earlier discussion, my service cannot pause for more than 10 sec as it would be result in time out for the clients that are connected to this service. What my understanding towards the checkpoint is taking my below settings as example, WAL size would never grow beyond 1MB because check point occurs after WAL file size reaches 1 MB. Please correct me if I am wrong. Settings: PRAGMA synchronous=NORMAL;",with default auto check point and page size = 1024 bytes.
On Thursday, 20 February 2014 4:21 PM, Richard Hipp <d...@sqlite.org> wrote: On Thu, Feb 20, 2014 at 7:08 PM, veeresh kumar <veeru...@yahoo.com> wrote: Hi, > I am using "PRAGMA >journal_mode=WAL;" with "PRAGMA synchronous=NORMAL;",with >default auto check point and page size = 1024 bytes. Since checkpoint >occurs automatically after every 1 MB, how much time it checkpoint operation >would take to complete as the database size grows large (range 1GB - 50 >GB). I understand that it depends on the hardrive, but on a very good >configuration, >will this operation ever exceeds 10 sec?Also during this operation, does the >database gets locked? > The time needed for a checkpoint depends much more on the size of the WAL file than on the size of the database. For a 1MB WAL file on modern hardware, I would think a checkpoint would require perhaps 50 to 100 milliseconds. Have you done measurements to see how long it takes on your system? The database cannot be written while a checkpoint is underway. But reads can run concurrently with a checkpoint. -- D. Richard Hipp d...@sqlite.org _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
